Aero recipe: document sampled forces and preserve exported quadrature - #1971
Open
peterdsharpe wants to merge 13 commits into
Open
peterdsharpe wants to merge 13 commits into
peterdsharpe wants to merge 13 commits into
Conversation
MeshToDomainMesh in cell_centroids mode turns each source cell into an interior query point and discards the cells. Any integral over those points (forces, area-weighted losses or metrics) then has no measure to weight by, and after SubsampleMesh the retained cells' measure weights are lost as well. Record cell_measures(mesh) (area times composed measure weights) on the interior under the reserved point_data key TARGET_QUADRATURE_MEASURE_KEY, aligned one-for-one with the centroids. Reject the key as a user target or as a pre-existing input field so it cannot be silently shadowed. Also validate the shape of the reserved per-cell measure-weights field and of tensor factors passed to compose_measure_weights: TensorDict only checks the leading dimension, so a (n_cells, 1) tensor was storable but broadcast wrongly against cell_areas.
…, drop query measure from saved outputs forces.py and infer.py said that force/moment coefficients from a subsampled vehicle surface "cover only the kept cells" and shrink by the kept-area fraction. That is no longer true: SubsampleMesh records each kept cell's inverse inclusion probability as a measure weight and Mesh.integrate multiplies by it, so the coefficients are unbiased (if noisy) estimates of the full-surface integrals. Correct the module docstring and the once-per-run warning, and pin the behaviour with a test on a subsampled closed surface. MeshToDomainMesh now records the query measure on the interior. It is in training-geometry units and would be stale after rescale_geometry, so attach_and_save drops it alongside the training-space targets.
Contributor
CODEOWNERS review mapCurrent for commit ⏳ @coreyjadams — 8 file(s)
⏳ @negin513 — 4 file(s)
⏳ @peterdsharpe — 6 file(s)
No CODEOWNER
Comment |
Collaborator
Author
|
/ok to test e39cd39 |
…dit-fix-1971-sept9 Signed-off-by: Peter Sharpe <peterdsharpe@gmail.com>
Signed-off-by: Peter Sharpe <peterdsharpe@gmail.com>
peterdsharpe
changed the title
Collaborator
Author
|
/ok to test a71c242 |
Signed-off-by: Peter Sharpe <peterdsharpe@gmail.com>
Signed-off-by: Peter Sharpe <peterdsharpe@gmail.com>
peterdsharpe
changed the title
Collaborator
Author
|
/ok to test fc52b33 |
peterdsharpe
marked this pull request as ready for review
September 19, 2026 20:57
peterdsharpe
requested review from
coreyjadams,
megnvidia and
negin513
as code owners
September 19, 2026 20:57
Contributor
 The PR is not safe to merge until the GLOBE visualization force integration preserves and consumes the corrected measure on subsampled prediction surfaces. Findings
SummaryThis PR replaces multiplier-style mesh weights with complete cell and point measures, propagates those measures through sampling and geometry operations, introduces explicit point-sample integration, and updates the external-aero documentation and inference contracts.
Reviews (1) · Last reviewed commit: "Merge unified measures into aero force d..." |
![greptile-apps[bot]](https://avatars.githubusercontent.com/in/867647?s=60&v=4){kind=small}
Comment on lines
+301
to
+302
| total_area = cell_measures(mesh).sum() | ||
| scale_measures(boundary, total_area / cell_measures(boundary).sum()) |
Contributor
There was a problem hiding this comment.
The visualization path subsamples prediction_mesh and stores the represented-area correction in cell_data["_effective_measure"]. Postprocessing then rebuilds pred_surface with empty cell data, and compute_surface_force_coefficients weights forces using geometric cell_areas. This discards the correction, so the reported predicted Cd/Cl/Cs shrink with the retained surface fraction. Preserve the effective measure on pred_surface and use cell_measures(surface_mesh) for integration.
Collaborator
Author
|
/ok to test f4475b3 |
This branch has not been deployed
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Stacked on #1970. Its current head (
e2ff58109108d67da4bb26eb8bb35a47b6cb045e) is merged into this branch. Relative to #1970, this PR changes only the force/inference documentation and warning, their regression tests, and the changelog.The aero recipe's force documentation and inference warning said subsampled coefficients shrink by the retained-area fraction. Mesh integration already uses complete effective measures, including sampling corrections. This PR documents that behavior and verifies that exported query measures remain usable in physical coordinates.
cell_measures(mesh). Exact Horvitz–Thompson unbiasedness requires correct inclusion probabilities, fixed field values, and a fixed physical moment origin. Approximate sampling, predictions depending on the sampled geometry, and sample-dependent centering can introduce bias; check convergence with surface resolution.No force-computation or sampling algorithm changes are included. These recipe contracts apply to all supported surface models.
Validation: all 33 force/inference tests and all applicable pre-commit hooks passed. The broader recipe suite had 185 passes, one skip and one failure in
test_sdf_normals_near_wall_use_face_normal. The identical signed-distance assertion failure reproduces on the pre-update #1971 head (22219d40) and #1970, independently of this change.